Implicit channel sounding for closed-loop transmission in MIMO-OFDM wireless networks

Abstract

Symbols are encoded in a transmitter of a wireless network that uses multiple-input and multiple-output (MIMO) orthogonal frequency division multiplexing (MIMO-OFDM) by partitioning the sub-carriers to be used for transmitting the symbol into a set of K groups of contiguous sub-carriers. For each group k of NK−1 modulation symbols Dk(1), Dk(2), . . . , Dk(NK−1), the modulation symbol Dk(n), for group k, is mapped to a corresponding mapped modulation symbols Xk(n) according to

Description

FIELD OF INVENTION

This invention relates generally to estimating channels, signal-to-noise ratios (SNR), and signal-to-interference-plus-noise ratios (SINR) for adaptive data transmission in wireless communication networks, and more particularly to channel sounding, channel estimation, and signal quality estimation for closed-loop adaptive modulation and coding techniques with multiple transmitter and multiple receiver antennas over multiple-input and multiple-output (MIMO) orthogonal frequency division multiplexing (MIMO-OFDM) networks.

BACKGROUND OF INVENTION

Multiple-input and multiple-output, orthogonal frequency division multiplexing (MIMO-OFDM) technologies are becoming increasing popular for wireless data communication networks. It is well-known that the spatial degrees of freedom, which MIMO techniques provide, increase data rates without the need for excess transmission bandwidth. OFDM also provides frequency-selective scheduling gains in a MIMO cellular network. Furthermore, the combination of MIMO with OFDM modulation provides high data rate capabilities over wider transmission bandwidths with improved reliability against time- and frequency-selective channel fading, multi-user diversity and interference in cellular wireless networks.

By varying the modulation format, channel code rate, transmission power and signaling duration, adaptive data transmission can exploit the variation of wireless channels to improve reliability performance and increase data throughput. Because MIMO-OFDM wireless networks are characterized by channel selectivity in space, time and frequency domains, conventional wireless networks, such as networks designed according to the IEEE 802.11n, IEEE 802.16e and 3GPP LTE standards, use channel sounding by multiplexing known pilot symbols (pilot tones) with unknown data symbols in an OFDM symbol, i.e., the pilot symbols and the modulation data symbols do not overlap. Thus, a receiver can estimate the space-time-frequency channel and feed back a quality metric for the channels to a transmitter. Then, the transmitter can adjust its transmission parameters, such as modulation format, channel code rate, transmission power and signaling duration to adapt the transmissions to the channels in a rapidly varying environment.

Unfortunately, multiplexing of known pilot symbols with unknown data symbols leads to a reduced number of the data symbols within each OFDM symbol, thereby significantly reducing the data rate per OFDM symbol. More importantly, because the number of pilot symbols needed to sound a highly frequency-selective fading channel is different from the number of pilot symbols needed to sound a frequency-flat fading channel, a priori multiplexing of pilot symbols and data symbols at a specific overhead either wastes the number of sub-carriers to sound a frequency-flat fading channel, and the overhead is insufficient to sound a highly frequency-selective fading channel. Also, because many different configurations are possible in multiplexing the pilot symbols and the data symbols within the sub-carriers of the OFDM network, the consequence is that design and implementation of conventional channel sounding (CS) methods are less flexible.

Implicit Pilot Symbols

In our related application Ser. No. 12/827,591, we first describe the basic principals of our implicitly embedding of pilot symbols in data symbols of resource blocks in MIMO-OFDM networks. That Application deals with the relatively simple case of an open-loop network where there is no feed back of channel state information from the receiver to transmitter. There the assumptions are that the channel coherence time is relatively long, e.g., for an entire resource block or more, and the channel coherence bandwidth is relatively constant. This is adequate for many networks wherein the configuration of the network and the environment is relatively static, e.g., indoor networks where the receivers in computers, laptops and wireless telephones generally move infrequently. In such networks, it is not important to adapt modulation and coding schemes to a rapidly varying environment.

Now, we want to deal with the more difficult case where the above assumptions are not true, and the environment does vary, as in networks operating outdoors, and where the transceivers are highly mobile.

SUMMARY OF THE INVENTION

The embodiments of the invention provide a method for sounding a channel for channel estimation and closed-loop data transmission by utilizing adaptive modulation and coding techniques for multiple-input and multiple-output (MIMO) orthogonal frequency division multiplexing (MIMO-OFDM) networks.

We use implicit channel sounding (ICS). The channel sounding is performed using a predetermined set of OFDM symbols. In each OFDM symbol designated for channel sounding, the transmitter does not transmit any explicit pilot symbols for channel estimation. Instead, these pilot symbols (pilot tones) are implicitly encoded in unknown modulated data symbols. As a result, there is no need for additional allocation of valuable resources such as transmission power, time and bandwidth to transmit explicit pilot symbols.

This leads to significant savings in energy, time and bandwidth resources, thereby improving the overall network power and spectral efficiency. Because all the modulated data symbols in our implicit ICS method encode the known pilot symbols, the invention enables us to estimate a frequency-selective OFDM channel more accurately, and adapt the data rate, modulation order, transmission power, and channel code rate to varying channels for closed-loop transmission.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic of a wireless network according to embodiments of the invention;

FIGS. 2-3 are block diagrams of OFDM symbols and implicit channel sounding according to embodiments of the invention;

FIG. 4 is a block diagram of processing a sub-carrier group according to embodiments of the invention;

FIG. 5 is a block diagram of an SINR computation processer according to embodiment of the invention;

FIG. 6 is a block diagram of processing a sub-carrier group for multiple antennas according to embodiments of the invention; and

FIG. 7 is a block diagram of an SINR computation processer according to embodiment of the invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

As shown in FIG. 1, the embodiments of our invention provide a method 125 for estimating channels in a multiple-input and multiple-output (MIMO) orthogonal frequency division multiplexing (MIMO-OFDM) network 120. In the network, there is one channel between each transmit and receive antenna pair. With closed-loop feedback, the receiver transmits channel state information (CSI) 150 to the transmitter. The channel estimation enables us to perform closed-loop adaptive modulation and coding techniques at the transmitter. The channels are estimated in receiver 122 using implicit pilot symbols 100 periodically transmitted by the transmitter.

Implicit Channel Sounding for Single-Stream OFDM Networks

As shown in FIG. 2, we consider an OFDM network that is described by the following network parameters:

• • the OFDM network bandwidth is W Hz,
  • the OFDM network sampling frequency is F_S samples per second,
  • the sub-carrier spacing is Δ_f Hz,
  • the number of sub-carriers (or the FFT size) is N_FFT,
  • the cyclic prefix length is N_CP samples,
  • the number of left-guard sub-carriers 201 is N_L,
  • the number of right-guard sub-carriers is N_R, and
  • the number of sub-carriers in an OFDM symbol 210 isN_U=N_FFT−N_L−N_R−1,ignoring a DC sub-carrier 220.

For an example IEEE 802.16e network operating over a channel with a 10 MHz bandwidth, W=10 MHz, F_S=11.2 M samples per second, Δ_f=10.9375 kHz, N_FFT=1024, N_CP=128 samples, N_L=92 sub-carriers, N_R=91 sub-carriers, and N_U=1024−92−91−1=840. Values of W, F_S, Δ_f, N_FFT, N_CP, N_L, N_R, and N_U can also be obtained for other wireless standards such as IEEE 802.11n and 3GPP LTE networks.

The available number of sub-carriers, N_U, is partitioned into a set of K contiguous groups 230, wherein each group includes

$N_K=\frac{N_U}{K}$ sub-carriers, where N_K is an integer.

Because OFDM networks typically experience frequency selective channel fading, the number of sub-carriers within a group, N_K, is selected such that N_K exceeds the normalized channel coherence bandwidth N_C

$N_C=\mathrm{round}\left(\frac{1}{5{\sigma }_{\mathrm{RMS}}{\Delta }_f}\right),$ where round(x) is a nearest integer to x, and σ_RMS is a root-mean-square (RMS) channel delay-spread. With the above value of N_C, we need N_K≦N_C so that we can reasonably approximate the channel over N_K consecutive sub-carriers as frequency-invariant.

A set of N_K−1 modulation symbols to be transmitted over the k^th group of N_K sub-carriers is D_k (1), D_k (2), D_k (N_K−1) 240. The transmitted symbols on these N_K sub-carriers are X_k(1), X_k(2), . . . , X_k(N_K) 250.

The following describes a mapping 260 of D_k(1), D_k (2), . . . , D_k(N_K−1) to X_k(1), X_k (2), . . . , X_k(N_K). Each of the N_K symbols in the k^th group is transmitted with an average energy of E_S(k). If E_S is the average energy of each of the N−1 transmitted mapped data symbols, then predetermined positive constants α₁ α₂ and β211 are respectively selected according to the following functions as

${\alpha }_1\left(k\right)=\sqrt{E_S\left(k\right)}$ ${\alpha }_2\left(k\right)=\sqrt{\left(N_K-1\right)E_S\left(k\right)},\mathrm{for}k=1\mathrm{to}K.\beta \left(k\right)=\sqrt{\frac{E_S\left(k\right)}{N_K-1}}$

Note that this formulation is different than the corresponding function in the related Application incorporated herein by reference. There, the mapping is over the symbols and sub-carriers of an entire resource block. In contrast, here as shown in FIG. 2 the mapping is over groups of sub-carriers of a single OFDM symbol. As an advantage, the mapping can adapt to carrying channel conditions on a symbol-per-symbol basis, whereas our previous mapping demands a relatively long channel coherence time.

Using the values α₁(k), α₂(k) and β(K), the ICS mapping rule for the k^th group is described in the following steps:

Step 1: Input the N_K−1 modulation symbols

• • D_k(1), D_k(2), . . . , D_k (N_K−1) 240;
  • Input the positive constants α₁ (k), α₂ (k) and β(k) 211; and

Step 2: Perform the ICS data mapping 260 according to

$m\left(k\right)=\frac{1}{N_K-1}\sum _{n=1}^{N_K-1}{D}_k\left(n\right)$ $X_k\left(n\right)={\alpha }_1\left(n\right)\left(D_k\left(n\right)-m\left(k\right)\right)+\beta \left(k\right)n=1,\dots ,N_K-1$ $X_k\left(N_K\right)={\alpha }_2\left(k\right)m\left(k\right),$ where m(k) is a sample-mean of the set of modulation symbols α₁, α₂ and β are predetermined positive constants described in detail below.

Note that this formulation is also different than the corresponding function in the related Application.

The output of the ICS mapping unit for the k^th group of N_K−1 modulation symbols are mapped modulation symbols X_k(1), X_k(2), . . . , X_k(N_K) 250. The above ICS mapping rule is repeated for all the K groups with the following input and output modulation symbols, For k=1 to K:

• • Input Constellation Symbols: D_k (1), D_k (2), . . . , D_k (N_K−1);
  • Input the positive constants α₁(k), α₂ (k) and β(k);
  • Perform Step 1 and Step 2 of the ICS data mapping rule; and
  • Output modulation symbols: X_k(1), X_k (2), . . . , X_k(N_K)

It is important to note that the ICS mapping rule described in Steps 1 and step 2 is valid for any choice of α₁ (k), α₂ (k) and β(k), and is not restricted to the exemplary choice of α₁(k), α₂(k) and β(k). r

The total number of modulation symbols inserted into the OFDM symbol is K(N_K−1)=N_U−K, and the number of sub-carriers is N_U. That is, the efficiency of the proposed ICS mapping invention is

$\frac{N_U-K}{N_U}=1-\frac{K}{N_U}$ per sub-carriers, and the overall efficiency, by taking into the number of left, right, and DC sub-carriers, is

$\frac{N_U-K}{N_{\mathrm{FFT}}}=\left(1-\frac{K}{N_U}\right)\frac{N_U}{N_{\mathrm{FFT}}}=\left(1-\frac{K}{N_U}\right)\left(1-\frac{N_L+N_R+1}{N_{\mathrm{FFT}}}\right).$

As an example, for the IEEE 802.16e network operating over 10 MHz channel bandwidth, we have N_FFT=1024, N_L=92 sub-carriers, N_R=91 sub-carriers, and the number of used sub-carriers is N_U=1024−92−91−1=840. With K=84 and N_K=840/84=8, the efficiency per used sub-carrier is

$1-\frac{84}{840}=9\text{/}10=90$ percent, whereas the overall efficiency per OFDM symbol is

$\frac{9}{10}\times \frac{840}{1024}=73.82$ percent.

Like in our previous Application, and in contrast with conventional channel sounding techniques, our ICS mapping does not require transmission of known pilot symbols. Because our ICS mapping uses all available sub-carriers for data transmission, while simultaneously performing channel sounding, there is a significant efficiency improvement in the number of sub-carriers that are actually used for data transmission.

Using our ICS technique, we now describe methods of estimating the frequency-domain OFDM channel and detection of individual modulation symbolsD₁(1),D₁(2), . . . , D₁(N_K−1), . . . , D_k(1)D_k(2), . . . , D_k(N_K−1).

Because the number of sub-carriers N_K in a group is selected to be less than the channel coherence bandwidth N_C, we assume that the channel within a group of N_K sub-carriers is constant. For the k^th group, the received signal on each of the N_K sub-carriers within a given OFDM symbol is given by

$\left[\begin{array}{c}{Y}_{k,1}\left(n\right)\\ Y_{k,2}\left(n\right)\\ ⋮\\ Y_{k,L}\left(n\right)\end{array}\right]=\left[\begin{array}{c}{H}_{k,1}\\ H_{k,2}\\ ⋮\\ H_{k,L}\end{array}\right]X_k\left(n\right)+\left[\begin{array}{c}{Z}_{k,1}\left(n\right)\\ Z_{k,2}\left(n\right)\\ ⋮\\ Z_{k,L}\left(n\right)\end{array}\right],n=1,\dots ,N_K,k=1,\dots ,K,$ where Y_k,l (n) is the received signal on sub-carrier n of the group k at receiver antenna l, H_k,l is the channel response of the group k at receiver antenna l, and Z_k,l (n) is the interference-plus-noise component on sub-carrier n of the group k at receiver antenna l. Upon stacking the elements Y_k,l(n), l=1, . . . , L, into one vectorY_k(n)=[Y_k,l(n) . . . Y_k,L(n)]^T,and similarly forh_k=[H_k,l . . . H_k,L]^T,

we can writeY_k(n)=h_kX_k(n)+Z_k(n),whereZ_k(n)=[Z_k,l(n)Z_k,L(n)]^T is the vector-valued noise-plus-interference on sub-carrier n of the group k, for k=1 to K.

Estimation of frequency-domain channel for the k^th group

${\hat{h}}_k=\frac{\sum _{n=1}^{N_K-1}{Y}_k\left(n\right)}{\beta \left(k\right)\left(N_K-1\right)}=h_k+\frac{\sum _{n=1}^{N_K-1}{Z}_k\left(n\right)}{\beta \left(k\right)\left(N_K-1\right)}.$

Demodulate the symbols within the k^th group

${\hat{D}}_k\left(n\right)=\frac{1}{{\alpha }_1\left(k\right)}\left(\frac{{\hat{h}}_k^H{Y}_k\left(n\right)}{{\hat{h}}_k^H{\hat{h}}_k}-\beta \left(k\right)\right)+\frac{1}{{\alpha }_2\left(k\right)}\frac{{\hat{h}}_k^H{Y}_k\left(N_K\right)}{{\hat{h}}_k^H{\hat{h}}_k},$ where n=1, . . . , N_K−1.

After the channels for each of the K groups are obtained, the overall frequency domain channel within the N_U used sub-carries is

$\left[\begin{array}{c}{\hat{H}}_1\left(n\right)\\ {\hat{H}}_2\left(n\right)\\ ⋮\\ {\hat{H}}_L\left(n\right)\end{array}\right]={\hat{h}}_{\lceil \frac{n}{N_K}\rceil },\mathrm{for}n=1\mathrm{through}{N}_U,$ where ┌x┐ is the ceiling of x and Ĥ_l(n) is the estimate of channel on sub-carrier n on receiver antenna l.

The above described embodiment assumes that all the N_U sub-carriers within a given OFDM symbol are utilized for data transmission. If only a smaller number of sub-carrier groups are employed, i.e., less than K groups, i.e., the number of groups is <K, we can still estimate the channel on each of those used sub-carrier groups, ĥ_k, as described above, and perform a frequency-domain interpolation process, such as linear, spline and windowing based interpolation algorithms, to estimate the channel on the remaining sub-carrier groups.

An estimate of the average noise power, per receiver antenna (channel), for the k^th group of N_K sub-carriers is computed as

$\begin{array}{c}{\hat{\sigma }}_k^2=\frac{1}{{\mathrm{LN}}_K}\sum _{n=1}^{N_K}||Y_k\left(n\right)-{\hat{h}}_k{\hat{X}}_k\left(n\right){||}^2\\ =\frac{1}{{\mathrm{LN}}_K}\sum _{n=1}^{N_K}||Y_k\left(n\right)-\frac{{\hat{h}}_k^H{Y}_k\left(n\right)}{{\hat{h}}_k^H{\hat{h}}_k}{\hat{h}}_k{||}^2.\end{array}$

Using the above estimated noise power, and an estimate of the channel ĥ_k on the k^th sub-group, the effective signal-to-noise ratio (SNR) for the k^th group of N_K sub-carriers is computed as

${\gamma }_k=\frac{{\hat{h}}_k^H{\hat{h}}_k}{{\hat{\sigma }}_k^2}=\sum _{l=1}^L\frac{|{\hat{H}}_{k,l}{|}^2}{{\hat{\sigma }}_k^2}.$

The estimated SNR values, γ₁, γ₂, . . . , γ_K, can now be utilized to determine an appropriate choice of the modulation order and channel code rate. If we use well-known channel abstraction techniques, such as mean mutual information per bit (MMIB), received bit information rate (RBIR), or exponentially effective SNR mapping (EESM) rule as specified for the IEEE 802.16m-08/004 r2 standard, in Evaluation Methodology Document (EMD), Srinivasan et al., Jul. 3, 2008, then a measure of the channel quality over the K groups is necessary.

Using γ₁, γ₂, . . . , γ_K, the effective SNR over the K groups with EESM rule is

$\begin{array}{c}{\gamma }_{\mathrm{Effective}}=-c\log \left(\frac{1}{K}\sum _{k=1}^K{e}^{-\frac{{\gamma }_k}{c}}\right)\\ =-c\log \left(\frac{1}{K}\sum _{k=1}^K{e}^{-\frac{{\hat{h}}_k^H{\hat{h}}_k}{c{\hat{\sigma }}_k^2}}\right)\\ =-c\log \left(\frac{1}{K}\sum _{k=1}^K{e}^{-\sum _{l=1}^L\frac{|{\hat{H}}_{k,l}{|}^2}{c{\hat{\sigma }}_k^2}}\right),\end{array}$ where c is a constant that depends on the modulation and coding employed. In practice, values of c are obtained a priori using the channel-level simulation techniques. When c is not known a priori, or when c takes very large negative values, the following metric can be used to represent the effective SNR over the K groups

$\begin{array}{c}{\gamma }_{\mathrm{Effective}}=\frac{1}{K}\sum _{k=1}^K{\gamma }_k=\frac{1}{K}\sum _{k=1}^K\frac{{\hat{h}}_k^H{\hat{h}}_k}{{\hat{\sigma }}_k^2}=\frac{1}{K}\sum _{k=1}^K\sum _{l=1}^L\frac{|{\hat{H}}_{k,l}{|}^2}{{\hat{\sigma }}_k^2}\\ ={\gamma }_{\mathrm{Avg}},\end{array}$ which is the SNR averaged over the K groups.

Using γ_Effective in conjunction with an estimated block error rate (or the frame error rate) performance curves corresponding to the additive white Gaussian noise (AWGN) channel, e.g., using a look-up table, enables us to determine the appropriate choice of modulation order and channel code rate.

Implicit Channel Sounding for MIMO-OFDM Networks

As shown in FIG. 3, we now describe MIMO-OFDM networks with multiple transmitter antennas. Here, the number of transmitter antennas is M_T and the number of antennas at the receiver is M_R.

Similar to the single transmitter antenna case, we first divide the available number of sub-carriers N_U into K contiguous groups, where each group comprises of

$N_K=\frac{N_U}{K}$ sub-carriers. It should be noted that K is again selected so that N_K is an integer. Next, the N_K sub-carriers in a given group are divided into M_T sub-groups in such a way that each sub-group contains

$N_{K,T}=\frac{N_K}{M_T}$ sub-carriers. Note that for a given M_T, N_K has to be selected in such a way that M_T divides N_K so that N_K,T becomes an integer. As an example, with M_T=2 transmitter antennas, K=84 and N_K=8, we have

$N_{K,T}=\frac{8}{2}=4.$

We denote by D_k,l(1), D_k,l(2), . . . , D_k,l(N_K,T−1) the N_K,T−1 modulation symbols assigned for k^th group on the l^th transmitter antenna. The ICS mapping rule with multiple transmitter antenna takesD_k,l(1),D_k,l(2), . . . , D_k,l(N_K,T−1)as the input and produces X_k,l(1), X_k,l(2), . . . , X_k,l(N_K,T), where X_k,l(n) is the modulation symbol that is transmitted on sub-carrier n assigned to k^th group on the l^th transmitter antenna.

Each of the N_K,T, symbols, X_k,l(1), X_k,l(2), . . . , X_k,l(N_K,T), in the k^th group corresponding to the l^th transmitter antenna is transmitted with an average symbol energy of E_S,l(k). Then, we define the following parameters for k=1 to N_K and l=1 to M_T,

${\alpha }_{1,l}\left(k\right)=\sqrt{E_{S,l}\left(k\right)}$ ${\alpha }_{2,l}\left(k\right)=\sqrt{\left(N_{K,T}-1\right)E_{S,l}\left(k\right)}$ ${\beta }_l\left(k\right)=\sqrt{\frac{E_{S,l}\left(k\right)}{N_{K,T}-1}}.$

Using the above values of α_1,l(k), α_2,l(k) and β_l(k), the ICS mapping rule for the k^th group with l^th transmitter antenna is described in the following steps:

Step 1: Input the N_K,T−1 modulation symbolsD_k,l(1),D_k,l(2), . . . , D_k,l(N_K−1)

• • Input the positive constants α_1,l(k), α_2,l(k) and β_l(k).

Step 2: Perform the ICS data mapping:

$m_l\left(k\right)=\frac{1}{N_{K,T}-1}\sum _{n=1}^{N_{K,T}-1}{D}_k\left(n\right)$ $X_{k,l}\left(n\right)={\alpha }_{1,l}\left(k\right)\left(D_{k,l}\left(n\right)-m_l\left(k\right)\right)+{\beta }_l\left(k\right)n=1,\dots ,N_{K,T}-1$ $X_{k,l}\left(N_{K,T}\right)={\alpha }_{2,l}\left(k\right)m_l\left(k\right)$

The output of the ICS mapping unit for the k^th group of N_K,T−1 modulation symbols to be transmitted on l^th transmitter antenna isX_k,l(1),X_k,l(2), . . . , X_k,l(N_K,T).

The above ICS mapping rule is repeated for all the K groups on all the M_T transmitter antennas with the following input and output modulation symbols for k=1 to K, and l=1 to M_T:

• • Input Constellation Symbols:D_k,l(1),D_k,l(2), . . . , D_k,l(N_K,T−1);
  • Input the positive constants α_1,l(k), α_2,l(k) and β_l(k),
  • Perform Step 1 and Step 2 of the ICS data mapping rule; and
  • Output modulation symbols: X_k,l(1), X_k,l(2), . . . , X_k,l(N_K,T).

It is important to note that the ICS mapping rule in Steps 1 and step 2 is valid for any choice of α_1,l(k), α_2,l(k) and ⊕_l (k), and is not restricted to the exemplary choice of α_1,l(k), α_2,l(k) and β_l(k) given preceding to Steps 1 and step 2.

It is also instructive to note that when M_T=1, this mapping rule reduces to the previous mapping rule.

The total number of modulation symbols inserted into the OFDM symbol with M_T transmitter antennas is

${\mathrm{KM}}_T\left(N_{K,T}-1\right)=N_U-{\mathrm{KM}}_T,$ and the number sub-carriers is N_U. That is, the efficiency of the ICS mapping for M_T transmitter antennas is

$\frac{N_U-{\mathrm{KM}}_T}{N_U}=1-\frac{{\mathrm{KM}}_T}{N_U}$ per used sub-carriers, and the overall efficiency, by taking into the number of left, right, and DC sub-carriers, is

$\frac{N_U-{\mathrm{KM}}_T}{N_{\mathrm{FFT}}}=\left(1-\frac{{\mathrm{KM}}_T}{N_U}\right)\frac{N_U}{N_{\mathrm{FFT}}}=\left(1-\frac{{\mathrm{KM}}_T}{N_U}\right)\left(1-\frac{N_L+N_R+1}{N_{\mathrm{FFT}}}\right)$

As an example, for the IEEE 802.16e network operating over 10 MHz channel bandwidth, we have N_FFT=1024, N_L=92 sub-carriers, N_R=91 sub-carriers, and the number of used sub-carriers is N_U=1024−92−91−1=840.

With K=84, M_T=2, and N_K=840/84=8, the efficiency per used sub-carrier is

$1-\frac{84\times 2}{840}=8\text{/}10=80$ percent, whereas the overall efficiency per OFDM symbol is

$\frac{8}{10}\times \frac{840}{1024}=65.61$ percent.

FIG. 4 shows the processing of each group at the receiver, FIG. 5 shows the process for computing the effective SINR, and FIG. 6 shows the processing of each group for multiple antennas, and FIG. 7 shows the process for computing the effective SINR using multiple antennas, all described in detail herein.

For the k^th group, the received signal on each of the N_K,T sub-carriers corresponding to transmitter antenna m within a given OFDM symbol is given by

$\left[\begin{array}{c}{Y}_{k,1}^m\left(n\right)\\ Y_{k,2}^m\left(n\right)\\ ⋮\\ Y_{k,L}^m\left(n\right)\end{array}\right]=\left[\begin{array}{c}{H}_{k,1}^m\\ H_{k,2}^m\\ ⋮\\ H_{k,L}^m\end{array}\right]X_{k,m}\left(n\right)+\left[\begin{array}{c}{Z}_{k,1}^m\left(n\right)\\ Z_{k,2}^m\left(n\right)\\ ⋮\\ Z_{k,L}^m\left(n\right)\end{array}\right],n=1,\dots ,N_{K,T},k=1,\dots ,K,m=1,\dots ,M_T,$ where Y_k,l^m(n) is the received signal on sub-carrier n of the group k at receiver antenna l when X_k,m(n) is transmitted from the transmitter antenna m, H_k,l^m is the channel response of the group k between the receiver antenna l and transmitter antenna m, and Z_k,l^m(n) is the interference-plus-noise component on sub-carrier n of the group k at receiver antenna l due to transmitter antenna m. Upon stacking the elements Y_k,l^m(n), l=1, . . . , L, into one vectorY_k^m(n)=[Y_k,l^m(n) . . . Y_k,L^m(n)]^T,and similarlyh_k^m=[H_k,l^m . . . H_k,L^m]^T,we can writeY_k^m(n)=h_k^mX_k,m(n)+Z_k^m(n),whereZ_k^m(n)=[Z_k,l^m(n) . . . Z_k,L^m(n)]^T is the vector-valued noise-plus-interference on sub-carrier n of the group k when transmitter antenna m is used.

We now estimate the vector-valued channels h_k^m for each transmitter antenna and for each group as follows, for k=1 to K, and m=1 to M_T.

Estimation of frequency-domain channel from transmitter m

${\hat{h}}_k^m=\frac{\sum _{n=1}^{N_{K,T}-1}{Y}_k^m\left(n\right)}{{\beta }_m\left(k\right)\left(N_{K,T}-1\right)}=h_k^m+\frac{\sum _{n=1}^{N_{K,T}-1}{Z}_k^m\left(n\right)}{{\beta }_m\left(k\right)\left(N_{K,T}-1\right)}.$

Demodulate the N_K,T−1 symbols transmitted by transmitter m within the k^th group

${\hat{D}}_{k,m}\left(n\right)=\frac{1}{{\alpha }_{1,m}\left(k\right)}\left(\frac{{\left({\hat{h}}_k^m\right)}^H{Y}_k^m\left(n\right)}{{\left({\hat{h}}_k^m\right)}^H{\hat{h}}_k^m}-{\beta }_m\left(k\right)\right)+\frac{1}{{\alpha }_{2,m}\left(k\right)}\frac{{\left({\hat{h}}_k^m\right)}^H{Y}_k^m\left(N_{K,T}\right)}{{\left({\hat{h}}_k^m\right)}^H{\hat{h}}_k^m},$ where n=1, . . . , N_K,T−1.

After the channels for each of the K groups from each of the M_T transmitter antennas are obtained, the overall frequency domain channel within the N_U used sub-carries is, for or n=1 to N_U,

$\left[\begin{array}{c}{\hat{H}}_1^m\left(n\right)\\ {\hat{H}}_2^m\left(n\right)\\ ⋮\\ {\hat{H}}_L^m\left(n\right)\end{array}\right]={\hat{h}}_{\lceil \frac{n}{N_K}\rceil }^m,$ where ┌x┐ is the ceiling of x and Ĥ_l^m(n) is the estimate of channel on sub-carrier n between receiver antenna l and transmitter antenna m.

An estimate of the noise variance on each of the L receiver antennas due to transmission from transmitter antenna m, for the k^th group of N_K sub-carriers, is

$\begin{array}{c}{\hat{\sigma }}_{k,m}^2=\frac{1}{{\mathrm{LN}}_{K,T}}\sum _{n=1}^{N_{K,T}}||Y_k^m\left(n\right)-{\hat{h}}_k^m{\hat{X}}_{k,m}\left(n\right){||}^2\\ =\frac{1}{{\mathrm{LN}}_{K,T}}\sum _{n=1}^{N_{K,T}}||Y_k^m\left(n\right)-\frac{{\left({\hat{h}}_k^m\right)}^H{Y}_k^m\left(n\right)}{{\left({\hat{h}}_k^m\right)}^H{\hat{h}}_k^m}{\hat{h}}_k^m{||}^2\end{array}$

Using the above estimated noise power, and an estimate of the channel ĥ_k^m on the k^th sub-group corresponding to transmitter antenna m, the effective signal-to-interference-plus-noise ratio (SINR) for the k^th group of N_K sub-carriers when to transmitter antenna m is used isγ_k^m=a_k^m(b_k^mγ_k,MRC^m+(1−b_k^m)γ_k,MMSE^m({circumflex over (σ)}_k,m²))where γ_k,MRC^m is the resulting SNR when maximum ratio combining (MRC) receiver is employed and γ_k,MMSE^m(λ) is the resulting SINR when linear minimum mean-square error (MMSE) receiver is employed with parameter λ. In the above equation, a_k^m is a positive constant that serves as a bias term, and 0≦b_k^m≦1.

One approach to obtaining appropriate values of a_k^m and b_k^m is using channel-level simulations. The SNR and SINR terms γ_k,MRC^m and γ_k,MMSE^m(λ) in the above equation are obtained as

${\gamma }_{K,\mathrm{MRC}}^m=\frac{{\left({\hat{h}}_k^m\right)}^H{\hat{h}}_k^m}{{\hat{\sigma }}_{k,m}^2}=\sum _{l=1}^L\frac{|{\hat{H}}_{k,l}^m{|}^2}{{\hat{\sigma }}_{k,m}^2}$ $\mathrm{and}$ ${\gamma }_{k,\mathrm{MMSE}}^m\left({\hat{\sigma }}_{k,m}^2\right)=\frac{1}{G_k^{\left({\hat{\sigma }}_{k,m}^2\right)}\left(m,m\right)}-1$ $\mathrm{where}$ $G_k^{\left(\lambda \right)}={\left[\lambda I_{M_T}+\left[\begin{array}{c}{\left({\hat{h}}_k^1\right)}^H\\ {\left({\hat{h}}_k^2\right)}^H\\ ⋮\\ {\left({\hat{h}}_k^{M_T}\right)}^H\end{array}\right]\left[{\hat{h}}_k^1{\hat{h}}_k^2\dots {\hat{h}}_k^{M_T}\right]\right]}^{-1}$ is of size M_T-by-M_T, I_MT is the M_T-by-M_T identity matrix, and G_k^(λ)(m,m) is the element correspond to m^th row and m^th column of G_k^(λ).

It is important to note that our SINR estimate γ_k^m is quite general and leads to the following special cases:

• • By setting λ=0 in G_k^(λ), a_k^m=1 and b_k^m=1, we obtain the SNR corresponding to MRC receiver;
  • By setting λ=0 in G_k^(λ), a_k^m=1 and b_k^m=0, we obtain the SINR corresponding to a linear zero-forcing receiver;
  • By setting λ={circumflex over (σ)}_k,m² in G_k^(λ), a_k^m=1 and b_k^m=0, we obtain the SINR corresponding to a linear MMSE receiver; and
  • By setting λ={circumflex over (σ)}_k,m² in G_k^(λ), and proper choice of a_k^m and b_k^m, we obtain an approximation of the SINR estimate for MIMO networks with non-linear maximum-likelihood (ML) receiver.
  • Thus, the SINR estimate γ_k^m is valid for both linear and non-linear receiver algorithms.

Upon using γ_k^m for k=1, . . . , K, and m=1, . . . , M_T, the effective SINR over the K groups with EESM rule is given by

$\begin{array}{c}{\gamma }_{\mathrm{Effective}}=-c\log \left(\frac{1}{K}\frac{1}{M_T}\sum _{m=1}^{M_T}\sum _{k=1}^K{e}^{-\frac{{\gamma }_k^m}{c}}\right)\\ =-c\log \left(\frac{1}{K}\frac{1}{M_T}\sum _{m=1}^{M_T}\sum _{k=1}^K{e}^{-\frac{a_k^m\left(b_k^m{\gamma }_{k,\mathrm{MRC}}^m+\left(1-b_k^m\right){\gamma }_{k,\mathrm{MMSE}}^m\left({\lambda }_k^m\right)\right)}{c}}\right),\end{array}$ where c is a constant that depends on the modulation and coding 501 employed. In practice, values of c are obtained a priori using the channel-level simulation techniques.

When c is not known a priori, or when c takes very large negative values, it is not uncommon to use the following metric to represent the effective SNR over the K groups:

$\begin{array}{c}{\gamma }_{\mathrm{Effective}}=\frac{1}{K}\frac{1}{M_T}\sum _{m=1}^{M_T}\sum _{k=1}^K{\gamma }_k^m\\ =\frac{1}{K}\frac{1}{M_T}\sum _{m=1}^{M_T}\sum _{k=1}^K\left[a_k^m\left(b_k^m{\gamma }_{k,\mathrm{MRC}}^m+\left(1-b_k^m\right){\gamma }_{k,\mathrm{MMSE}}^m\left({\lambda }_k^m\right)\right)\right]\end{array}$

Using γ_Effective in conjunction with the (block or frame) error rate performance curves corresponding to the additive white Gaussian noise (AWGN) channel enables us to determine the appropriate choice of modulation order and channel code rate.

Implicit Channel Sounding for Multi-User MIMO-OFDM Networks

Above, we described ICS between a single transmitter and a single receiver utilizing MIMO-OFDM techniques. This approaches can be extended to the case of multiple users.

When K_User users (transceivers) are present in the network, each with M_T transmitter antennas, we do the following operations to perform ICS.

Similar to the single user case, we partition the available number of sub-carriers, N_U, into K contiguous groups, where each group comprises of

$N_K=\frac{N_U}{K}$ sub-carriers. It should be noted that K is selected in such a way that is an integer.

From the available K sub-carrier groups, we assign K_j groups to user j. Because the total number of sub-carrier groups assigned to a given user cannot be negative, we have K_j≧0. Also, the total number of sub-carrier groups assigned to all the users cannot exceed the available sub-carrier groups, K. That is, we have

$\sum _{j=}^{K_{\mathrm{User}}}{K}_j\le K.$ As an example, for the choice of K=84 we can support up to 84 users, each using one group of 10 sub-carriers. On the other hand, we can also support 14 users each using 6 sub-carrier groups (or, a total of 6×10=60 sub-carriers per user).

After the sub-carriers groups are assigned to the individual users (according to their respective data rate requirements), each user performs the following:

The N_K sub-carriers in a given group for a given user are partitioned into M_T sub-groups in such a way that each sub-group contains

$N_{K,T}=\frac{N_K}{M_T}$ sub-carriers. Note that for a given M_T, N_K has to be selected in such a way that M_T divides N_K so that N_K,T becomes an integer. As an example, with M_T=2 transmitter antennas, K=84 and N_K=8, we have

$N_{K,T}=\frac{8}{2}=4.$

Although the invention has been described by way of examples of preferred embodiments, it is to be understood that various other adaptations and modifications may be made within the spirit and scope of the invention. Therefore, it is the object of the appended claims to cover all such variations and modifications as come within the true spirit and scope of the invention.

Claims

1. A method for encoding a symbol in a transmitter of a wireless network, wherein the network uses multiple-input and multiple-output (MIMO) orthogonal frequency division multiplexing (MIMO-OFDM), comprising the steps of: partitioning the sub-carriers to be used for transmitting the symbol into a set of K groups of contiguous sub-carriers, and for each group further comprising the steps: providing a set of NK−1 modulation symbols Dk(1), Dk(2), . . . , Dk(Nk−1), wherein there is one modulation symbol for each sub-carrier in the group;mapping each modulation symbol Dk(n), of group k, to a corresponding mapped modulation symbols Xk(n) according to

2. The method of claim 1, wherein ES(k) is the average energy of each mapped modulation symbol for group k, and, for each of the K groups, the constants α1(k), α2(k) and β(k) 211 are respectively selected according to the following functions as

3. The method of claim 1, wherein a number of sub-carriers in each group is selected to be less than the channel coherence bandwidth.

4. The method of claim 1, further comprising: estimating a channel between the transmitter and a receiver based on a received signal corresponding to the symbol.

5. The method of claim 2, wherein a number of groups used is less than K, and the estimating uses frequency-domain interpolation process to estimate the channels in the unused sub-carriers.

6. The method of claim 1, further comprising: estimating an average noise power per channel for each group.

7. The method of claim 1, further comprising: estimating an average effective signal-to-noise ratio and error rate for the K groups, and further comprising:selecting a modulation order and channel code rate to transmit the symbol.

8. The method of claim 1, wherein the transmitter includes MT antennas, and further comprising: partitioning each sub-group into MT sub-groups; andperforming the providing and mapping for each sub-group.

9. The method of claim 7, further comprising: estimation of a per-stream signal-to-interference-plus-noise ratio for a linear zero-forcing receiver with multiple spatial data streams; andselecting, for each spatial data stream, a modulation order and a channel code to transmit the spatial data stream.

10. The method of claim 7, further comprising: estimation of a per-stream signal-to-interference-plus-noise ratio for a linear minimum-mean-square error receiver with multiple spatial data streams; andselecting, for each spatial data stream, a modulation order and a channel code to transmit the spatial data stream.

11. The method of claim 7, further comprising: estimation of a per-stream signal-to-interference-plus-noise ratio for a non-linear maximum-likelihood receiver; andselecting, for each spatial data stream, a modulation order and a channel code to transmit the spatial data stream.

12. The method of claim 7, wherein a number of users is KUser, and a number of transmitter antennas per-user is MT, further comprising: partitioning an available number of sub-carriers NU, into K contiguous groups, wherein each contiguous group includes of